1 Taxonomic and trait macroinvertebrate patterns across hydrological connectivity in

2 temperate, subtropical, Mediterranean, and arid river floodplains

3 Belinda Gallardoa,b*, Sylvain Dolédecc, Amael Paillexb,d, David B. Arscotte, Fran Sheldonf, Florencia Zillig,

4 Sylvie Mérigouxc, Emmanuel Castellad and Francisco A. Comína

5

6 a Applied and Restoration Ecology Group, Pyrenean Institute of Ecology (IPE-CSIC), Avda. Montañana

7 1005, 50192 Zaragoza (Spain)

8 b Aquatic Ecology Group, Department of Zoology, University of Cambridge, Downing St. CB2 3EJ

9 Cambridge (U.K.)

10 c Université Lyon 1; CNRS, UMR 5023, LEHNA, Biodiversité des Ecosystèmes Lotiques, F-69622,

11 Villeurbanne (France)

12 d Laboratory of Ecology and Aquatic Biology, University of Geneva, Route de Rize 7, 1207 Carouge

13 (Switzerland)

14 e Stroud Water Research Center, Avondale, PA 19311 (USA)

15 f Australian Rivers Institute, Griffith University, Nathan, Qld. 4111 (Australia)

16 g Instituto Nacional de Limnologia (UNL-CONICET), Ciudad Universitaria S/N, Ruta 168, Santa Fe 3000

17 (Argentina)

18 *Correspondence to:

19 Dr Belinda Gallardo

20 Current address:

21 Aquatic Ecology Group, Department of Zoology, University of Cambridge

22 Downing St. CB2 3EJ Cambridge (UK)

23 e-mail: [email protected] / [email protected]

24 Phone: +44 01223336617 / Fax : +44 0122336676

25

1

26 Abstract

27 Despite the wide recognition that benthic macroinvertebrates respond to changes in hydrological

28 connectivity within floodplain ecosystems, no generalizations about patterns in community structure

29 and functionality across large scales and different climates have been yet established. Using information

30 from six large rivers located in four different climate regions (subtropical, temperate, Mediterranean

31 and arid), we compared benthic macroinvertebrate responses along lateral gradients of hydrological

32 connectivity. We tested hypotheses related to differences among climate regions and to similar

33 hydrological constraints under any climate. First, multivariate ordinations demonstrated a higher

34 overlap of trait community composition (55.7% variance explained by the first two axes) than taxonomic

35 composition (19.2%) among floodplains, suggesting high inter-climate trait stability. Second, across a

36 gradient of lateral connectivity, Linear Mixed Effect (LME) models supported 7 out of 15 hypotheses,

37 which evidenced remarkably consistent macroinvertebrate patterns in floodplains regardless of climate

38 regime. Taxon and trait richness were positively related and peaked in sites with intermediate

39 hydrological connectivity. Richness of Ephemeroptera, Trichoptera, and Plecoptera peaked in highly

40 connected sites, while richness of Coleoptera, Odonata and Heteroptera was higher in the most

41 fragmented sites. Our expectations related to the feeding structure of macroinvertebrates (e.g.

42 shredders and scrapers more abundant in connected channels, and predators and deposit feeders in

43 fragmented sites) were better supported by observations than those regarding life history (e.g.

44 plurivoltinism and short life spans in connected channels). This difference was related to the influence of

45 extended periods of fragmentation (i.e. hydrological disconnection) as a complementary disturbance

46 factor to flooding in the filtering of traits. Trait similarities and dissimilarities found in this study suggest

47 that large-scale filters do operate on communities resulting in apparently different trait combinations in

48 temperate and Mediterranean floodplains when compared to arid and subtropical environments. This

49 latter finding needs to be considered to formulate ecologically meaningful a priori predictions.

50 Keywords: River Habitat Templet, Paraná River, Tagliamento River, Rhône River, Ebro River, Murray

51 River, Cooper Creek, aquatic macroinvertebrates

52

2

53 Introduction

54 Connectivity and fragmentation are fundamental concepts in ecology influencing the distribution of

55 organisms within an ecosystem (e.g. Moilanen and Nieminen 2002). Connectivity allows organisms to

56 move from an area to another, it increases the resilience of an ecosystem towards disturbance and

57 supports key ecosystem functions (e.g. Moilanen and Nieminen 2002). River floodplains comprise a

58 complex mosaic of interconnected aquatic and terrestrial patches that are arranged along a gradient of

59 lateral hydrological connectivity (e.g. Ward and Stanford 1995). This gradient of connectivity extends

60 from water bodies with a high frequency of connectivity, either due to their position adjacent to the

61 main channel (i.e. connected channels) or by being connected to the main channel at high flow levels

62 (e.g. water bodies that range in distance from the river channel), to those fragmented water bodies that

63 are only inundated under the most extreme floods (e.g. Amoros and Roux 1988). Since the first

64 definition of the Flood Pulse Concept in tropical environments (Junk et al. 1989) and its extension to

65 temperate areas (Tockner et al. 2000), there has been wide recognition that floodplain aquatic

66 communities respond to changes in hydrological connectivity in a predictable way (see the reviews of

67 Junk and Wantzen 2004, Petts and Amoros 1993). Nevertheless, to our knowledge no generalizations

68 about benthic macroinvertebrate structure and functionality across large scales and different climates

69 have been yet proposed. The lack of a general understanding on how benthic macroinvertebrate

70 communities respond across the hydrological gradient in floodplains may be partly due to a lack of

71 comparable field observations concerning their distribution from a range of rivers covering a broad

72 gradient of climatic and hydrological conditions (such as in Minshall et al. 1983).

73 In an attempt to provide a framework for comparing ecological processes and responses between

74 different river systems, a number of authors have undertaken river classifications from different regions

75 (e.g. Kennard et al. 2010, Poff and Ward 1989). Broadly, these classifications have identified groups of

76 rivers that range from perennial rivers and streams, characterised by consistent base flow and

77 predictable flood patterns –and often represented in temperate regions—to intermittent rivers and

78 streams that may have variable periods of cease to flow but can flood predictably. This latter case

79 concerns many Mediterranean and subtropical systems (e.g. Bonada et al. 2007, Depetris 2007),

80 whereas unpredictable floods occur in many arid and semi-arid systems (e.g.Kennard et al. 2010). Such

3

81 differences in climate and hydrology should result in different spatio-temporal floodplain conditions

82 yielding differences in taxonomic structure of aquatic communities (e.g. Poff and Ward 1990, Puckridge

83 et al. 1998). Owing to different taxonomic representations across broad regional scales, it can be

84 difficult to get a clear picture of similarities or general patterns if just relying on taxonomic comparisons.

85 In contrast, species traits across broad regions may provide insight into general patterns related to

86 hydrological connectivity, since traits aggregate biological information shared among taxonomically

87 different taxa. Therefore, while large differences in taxonomic composition are expected for rivers from

88 distinct climate or biogeographic regions, similarity in trait response across the hydrological gradient is

89 anticipated (e.g. Charvet et al. 2000).

90 Although many authors have reported a significant response of benthic macroinvertebrates to a

91 gradient of lateral connectivity, most of the available literature focuses on temperate European rivers,

92 such as the Rhône (e.g. Amoros and Bornette 2002, Castella et al. 1984, Paillex et al. 2007), the Danube

93 (e.g. Reckendorfer et al. 2006, Tockner et al. 1999, Ward et al. 1999) and the Rhine (e.g. Ward et al.

94 1999). Fewer studies exist in Mediterranean (but see Bonada et al. 2006, Gallardo et al. 2008), arid (but

95 see Sheldon et al. 2002, Sheldon and Thoms 2006), or subtropical rivers (but see Arrington and

96 Winemiller 2006, Marchese and Ezcurra de Drago 1992, Zilli et al. 2008). According to these and other

97 publications, EPT richness (Ephemeroptera, Plecoptera and Trichoptera) usually peaks in habitats with

98 high hydrological connectivity because of the high dissolved oxygen requirement and perhaps lower

99 thermal tolerance of associated species (Usseglio-Polatera and Tachet 1994). In contrast, COH richness

100 (Coleoptera, Odonata and Heteroptera) tends to peak in sites most disconnected from main channels

101 (e.g. Arscott et al. 2005, Bonada et al. 2006, Skern et al. 2010). Species richness and diversity usually

102 peak under intermediate hydrological disturbance conditions in accordance with the Intermediate

103 Disturbance Hypothesis (Connell 1978). Because of intense abiotic constraints, we might expect low trait

104 richness and diversity under high hydrological connectivity, while stability in intermediately connected

105 and fragmented channels may foster biotic interaction, thus trait richness (Statzner et al. 2004).

106 According to the River Habitat Templet (RHT, Townsend and Hildrew 1994), we might expect higher

107 proportions of small-sized individuals, associated with plurivoltinism and short life spans in the main

108 river channel submitted to frequent flooding. It is worth noting that highly variable flow regimes and

4

109 prolonged periods of extended hydrological fragmentation in intermittent rivers may reverse this

110 pattern, with small short-lived plurivoltine organisms dominating fragmented floodplain water bodies

111 that eventually dry up (Sheldon et al. 2010). In addition, gradual changes in the feeding structure along a

112 gradient of lateral connectivity can be anticipated because of differences in the availability of food

113 sources. In particular, a prominence of shredders, scrapers, and filter feeders can be predicted in

114 connected sites; whereas a higher proportion of individuals adapted to resist desiccation, together with

115 a higher frequency of predators and deposit feeders should be expected in fragmented habitats.

116 Because different biological traits confer clear trade-offs – mainly body size and life history

117 characteristics (for instance predators are generally large and univoltine, small organisms are generally

118 plurivoltine and with a short life span) — we should expect associated traits to dominate under the

119 same hydrological conditions. Most of these predictions have been confirmed in temperate rivers, while

120 significantly less information on trait patterns of macroinvertebrate is available for non-temperate

121 climate regions, and the studies that compare different regions are even rarer (but see Bonada et al.

122 2007). These gaps in knowledge are noteworthy since anthropogenic activities continue to alter the

123 natural hydro-geomorphic function of large floodplains with most consequences to aquatic communities

124 still unknown (Wantzen et al. 2008).

125 Using the available information from six large floodplain rivers located in four different climate regions

126 (subtropical, temperate, Mediterranean and arid) we compared patterns in benthic macroinvertebrates

127 across hydrological connectivity gradients. We tested two types of hypotheses: those associated with

128 differences among climate regions and those testing a common response of macroinvertebrate

129 communities to similar hydrological constraints under any climate (see Table 1 for a summary of

130 predictions). Ultimately, this study is aimed to provide a comprehensive comparison of aquatic

131 community patterns across hydrological gradients under different climate settings, thus increasing our

132 understanding of river floodplain dynamics and interactions in a wide context.

5

133

134 Methods

135 Case study floodplains

136 We considered six different large rivers and their floodplains as representative of the subtropical

137 (Paraná River), temperate (Tagliamento and Rhône Rivers), Mediterranean (Ebro River) and arid (Cooper

138 Creek, Diamantina, Murray and Darling Rivers) climates (Peel et al. 2007) (Fig. 1). The selection of river

139 floodplains was based on the availability of comparable data with respect to number and distribution of

140 sampling sites (i.e. across floodplain connectivity gradients), sampling methodology, and level of

141 taxonomic resolution.

142 The subtropical Paraná River is characterized by high temperatures (> 15ºC) throughout the year and

143 marked rainfall seasonality, which drives the predictable river discharge: high during summer and low

144 during winter and early spring (Depetris 2007). The Paraná River has a 900 km long floodplain, 30% of

145 which is occupied by a high number of relatively shallow (< 5 m depth) water-bodies of variable shape

146 and size, permanent and ephemeral, directly or indirectly connected to the river channel (Drago 1989).

147 Only extraordinary flooding events related to ‘El Niño Southern Oscillation’, completely inundate

148 habitats located in the floodplain and connect water bodies that are seldom affected by smaller events

149 (Depetris et al. 1996, Drago et al. 2003).

150 Temperate rivers are generally distinguished by permanent discharges and relatively predictable floods

151 throughout the year, which reflect a high and relatively constant rainfall with very infrequent drought

152 periods. Among temperate rivers, the Upper Rhône (from SW Switzerland to SE France) develops a great

153 diversity of floodplain habitats caused by river dynamics despite being moderately-to-highly affected by

154 dams, water abstraction, and lateral flood protection measures (Olivier et al. 2009). Its hydrologic

155 regime is nivo-glacial, with annual floods normally occurring in spring and low flows in late autumn to

156 winter (Olivier et al. 2009). In contrast, the Tagliamento (NE Italy) is one of the last morphologically

157 intact rivers draining the European Alps, characterised by a flashy pluvio-nival hydrological regime, with

158 highest discharges during fall and slightly lower magnitude floods occurring again in the spring (Arscott

159 et al. 2005). Although it falls within the alpine to Mediterranean climatic region (Peel et al. 2007),

6

160 Tockner et al. (2003) considered the Tagliamento River a reference ecosystem not only to the Alps but

161 also to other large temperate rivers. We hereafter refer to the Tagliamento as temperate for clarity.

162 Mediterranean rivers are characterized by the seasonality and variability of rainfall and temperature,

163 which results in highly variable discharges, torrential floods, and severe droughts (Bonada et al. 2007,

164 Gasith and Resh 1999). The Ebro River in NE Spain is an important river in the Mediterranean basin due

165 to its very long main stem channel and its relatively high discharge. Flow regulation in the Ebro River has

166 resulted in partial fragmentation of the floodplain (Cabezas et al. 2009a), although flooding still occurs

167 between October and March, strongly structuring aquatic communities (Gallardo et al. 2009b).

168 Arid rivers are typically characterised by extreme variability in flow (Puckridge et al. 1998) with

169 considerable periods of time spent as fragmented waterholes both within the main channel and across

170 the floodplain. Combined with geomorphic complexity, this provides a mosaic of hydrological

171 connectivity both in space and time (Sheldon et al. 2002), which makes them different from subtropical,

172 temperate and Mediterranean permanent rivers. The Cooper Creek and Diamantina River (“Cooper”

173 hereafter as they were considered together) drain the Lake Eyre Basin in central Australia. They are

174 characterized by extreme hydrological variability and a braided network of channels and tributaries

175 lowly affected by human activities (Bunn et al. 2003, Knighton and Nanson 1994). In comparison, in the

176 Murray-Darling system (“Murray” hereafter), water resource development has reduced the frequency

177 and magnitude of small to medium flood events, thereby decreasing hydrological connectivity across

178 the floodplain whereas the presence of in-channel weirs has increased hydrological connectivity along

179 the main channel waterholes (Sheldon and Thoms 2006, Walker and Thoms 1993). Because of the

180 intermittent nature of arid rivers we might expect macroinvertebrate life history patterns summarized in

181 Table 1 to notably differ from the rest of the rivers investigated. In arid rivers, waterholes in the main

182 river channel are the most stable habitat because of their more frequent flow, whereas connection

183 frequency decreases in the more fragmented habitats. Consequently, life-history traits conferring

184 organisms the ability to cope with frequent flooding (e.g. small body size, plurivoltinism, and short life

185 span, Table 1) may also occur in organisms adapted to temporary habitats and therefore be more

186 prominent in fragmented waterholes of arid rivers.

7

187 Water bodies sampled in the six river floodplains were divided into three homogeneous habitat types

188 that could be easily arranged along a gradient of lateral connectivity from high (i.e. the main river and

189 secondary channels being connected both upstream and downstream), intermediate (i.e. water bodies

190 partially connected, usually downstream), to low (i.e. fragmented water bodies with no upstream or

191 downstream connection) (Table 2). For all studies we considered samplings conducted during the

192 summer when fragmentation between the main river channel and floodplain habitats is highest and

193 marked differences across hydrological categories might be expected, as the summer time favours the

194 development of macroinvertebrate populations. More information on sites selected in each river and

195 methods used to sample macroinvertebrates are available in Table 2.

196

197 Macroinvertebrate sampling

198 Benthic macroinvertebrates were collected using a variety of methods (see Table 2 column Benthic

199 macroinvertebrates sampling). Consequently, differences were detected among sampling

200 methodologies in macroinvertebrate abundance and richness (e.g. higher in Hess samples of the

201 Tagliamento River, Kruskal-Wallis test, P< 0.001), though we could not separate the effect of sampling

202 method and river. In this sense, methodological differences are likely to affect among floodplain

203 analyses, e.g. clustering of rivers using the same method in ordination plots. All taxonomic

204 identifications were converted to genus level with a few exceptions such as family level for the

205 Chironomidae and order for Oligochaeta (see Appendix 1); these taxonomic groupings have been shown

206 as suitable for comparing the general structure of communities, i.e. in terms of taxa and traits (Dolédec

207 et al. 1999, Gayraud et al. 2003). As a result, a total of 228 taxa were included in subsequent analyses

208 (Appendix 1). Abundance data was converted to densities (number of individuals/m2), and densities

209 were log10 (x+1) transformed to reduce numerical disparities. Total richness, Shannon diversity, and EPT

210 (Ephemeroptera, Plecoptera, and Trichoptera) and COH (Coleoptera, Odonata, and Heteroptera)

211 richness were calculated and compared among rivers (see Data analysis section).

212 A total of 11 benthic macroinvertebrate traits were extracted from Tachet et al. (2010) to calculate trait

213 composition, richness and diversity (Table 3). These biological traits described different aspects of

214 organismal biology including body size, reproduction technique, resistance forms, and feeding habits.

8

215 For 38 out of 228 taxa not included in Tachet et al. (2010), scores were assigned based on average

216 values derived from each taxonomic family, or the closest taxon (see Appendix 2 for uncoded taxa and

217 families/taxa used as reference). This was the best option in the absence of individual trait information,

218 though we acknowledge the subsequent loss of accuracy by this approximation, and the possibility of

219 real scores being different from those used in this study. Traits were quantified using a “fuzzy coding”

220 technique (see Chevenet et al. 1994), a well-established method of coding (e.g. Gayraud et al. 2003,

221 Usseglio-Polatera et al. 2000) that involves the assignment of an affinity score, from 0 to 3, of each

222 taxon to each category for a given trait (see more about macroinvertebrate trait coding and analysis in

223 Usseglio-Polatera et al. 2000). This approach thus acknowledges trait variability, which often occurs

224 across regions or among different life stages. We further described the trait composition of communities

225 by multiplying the frequency of each category per trait by the relative log-transformed densities of

226 species at a given site. The resulting trait-by-site array contained the relative abundance of each

227 category per trait in each site. Finally, trait richness was calculated as the number of trait categories

228 present in a site, and trait diversity as the Rao diversity coefficient, using the methodology developed by

229 Champely and Chessel (2002) which incorporates both the relative abundance of species and a measure

230 of the pairwise functional differences between species (Botta-Dukat 2005).

231

232 Data analysis

233 Multivariate analyses (Correspondence Analysis, CA, and Fuzzy Correspondence Analysis, FCA) were

234 used to assess taxonomic and trait differences among river floodplain systems. Because CA favours rare

235 taxa (Hill 1974), those taxa with < 3 individuals or appearing in < 3 sites were removed. To further

236 investigate differences among hydrological categories (High, Intermediate and Low), we performed a

237 ‘between-class’ CA (Dolédec and Chessel 1989). To account for the effect of hydrological connectivity

238 only, we first compensated the effect of rivers using a ‘within class’ CA with ‘river identity’ as a factor,

239 and then performed a between-class CA on the residual table using ‘hydrological connectivity’ as the

240 instrumental variable (ter Braak 1986). This particular case of nested canonical correspondence analysis

241 first accounted for the partition of samples into the 6 different rivers and focused on structures that

242 were common to all groups (Dray et al. 2007). The significance of the between-class variance was

9

243 further tested against simulated values obtained after 1,000 permutations of the rows of the

244 invertebrate composition table for both taxon and trait composition.

245 Linear Mixed Effect models (LME) were used to extract common patterns of benthic invertebrate

246 communities across lateral connectivity gradients. This robust statistical technique is increasingly

247 exploited for its capacity to avoid problems related to the lack of independence of ecological

248 observations (Bolker et al. 2009). River identity was included as a random factor, while hydrological

249 connectivity (High, Intermediate and Low) was used as fixed effects. A total of 15 models were

250 developed to test our initial predictions, summarized in Table 1. Because results from models using

251 traits were uneven, and sometimes not significant, Kruskal-Wallis analyses of variance were performed

252 by river (Appendix 3). To avoid false significant results by chance due to multiple comparisons, a Keppel

253 correction of the significant threshold was applied, which accounted for the number of degrees of

254 freedom in the model (Keppel 1991).

255 Statistics and graphical outputs were computed with the libraries ade4 (Chessel et al. 2004, Dray and

256 Dufour 2007, Dray et al. 2007), vegan (Oksanen et al. 2007) and nlme (Pinheiro et al. 2010); all of which

257 were implemented in R CRAN (R Development Core Team 2011).

258

259 Results and Discussion

260 Benthic macroinvertebrate patterns across climate regions

261 Taxonomic structure among rivers

262 According to multivariate ordinations, the taxon composition of subtropical, temperate, Mediterranean,

263 and arid floodplains strongly differed, as might be expected considering their intrinsic climatic,

264 geological, geomorphological and hydrological features. The first three axes of the correspondence

265 analysis (CA) utilizing the taxonomic composition dataset accounted for only 19.2% of the total

266 variability, illustrating considerable taxonomic heterogeneity across climate regions. The first and

267 second CA axes separated the subtropical Paraná River on the upper right corner from the other five

268 rivers, and were positively related to non-insect taxa such as Bivalvia (e.g. Pisidium sp., Eupera sp.,

269 Musculium sp.) and Ephemeroptera (e.g. Campsurus sp.) (Fig. 2). Furthermore, the Paraná River reached

270 an average 86.2% of non-insect species, which is higher than the values obtained for temperate (38.8%),

10

271 Mediterranean (38.7%) and arid (41.99%) river floodplains (Kruskal-Wallis test, df=5, Chi2=71.05,

272 p<0.001). Other authors have found the main channel of the Paraná River to be highly dominated by

273 non-insects – particularly the endemic oligochaete Narapa bonettoi — due to instability of the sandy

274 substrate, while bivalves and gastropods were more frequently observed in floodplain lakes (Marchese

275 et al. 2002). A significant change in aquatic insect diversity with latitude has been also observed before,

276 as illustrated by greater richness of Ephemeroptera and Plecoptera with increasing latitudes, which was

277 related to their adaptation to cool waters (Pearson and Boyero 2009). Accordingly, in the present study

278 these two groups composed a greater percentage of the community in temperate and Mediterranean

279 river floodplains (average 10.9%) than in subtropical (8.6%) and arid (4.5%) systems located at lower

280 latitudes (Kruskal-Wallis test, Chi2= 77.68, df=5, p<0.001).

281 The first and second CA axes were negatively related to the abundance of insects, which accounted for

282 more than 50% of all individuals in temperate, Mediterranean, and arid river floodplains, and included

283 Coleoptera (e.g. Graptodytes sp., Noterus sp.), Odonata (e.g. Anax sp., Sympetrum sp.), Heteroptera

284 (e.g. Nepa sp., Hydrometra sp.), Diptera (e.g. Ptychopteridae, Blepharicera sp.), Plecoptera (e.g. Isoperla

285 sp., Chloroperla sp.) and Trichoptera (e.g. Philopotamus sp, Rhyacophila sp.). The third axis (not shown)

286 best discriminated arid rivers, characterized by the presence of particular Coleoptera (e.g. Hyderodes

287 sp., Cybister sp.), Gastropoda (e.g. Thiara sp., Notopala sp.) and Crustacea (e.g. Paratya sp., Caridina

288 sp.), all of them being found in arid rivers only. The between-class analysis revealed that differences

289 among hydrological connectivity categories accounted for only 4.1% of the initial CA variance, a low but

290 significant value (P< 0.001).

291

292 Trait structure among rivers

293 The first three axes of a fuzzy correspondence analysis (FCA) performed on trait composition explained

294 55.7% of the total variance, almost three-fold higher than the variance explained by the taxonomically

295 derived CA. In contrast to the taxonomic CA, the six river floodplains were less distinct in the trait-based

296 ordination, though the Paraná River was again clearly separated from the others towards the right

297 corner (Fig. 3). Traits positively related to the first FCA axis and thereby describing the Paraná alone,

298 included resistant forms (e.g. cells against desiccation, cocoons), large body size (between 4 and 8 cm),

11

299 asexual reproduction or deposition of free isolated eggs, tegument or aerial vesicle respiration, and

300 feeding on suspended material or organic deposits. Again, these results are consistent with other

301 studies of the Paraná River. For example, organic matter inputs, greatly influenced by flood pulses, have

302 been identified as a primary determinant of benthic macroinvertebrates in the Paraná River (Zilli et al.

303 2008). The body size of benthic macroinvertebrates in this river floodplain, probably driven by

304 dominance of bivalves and gastropods, was indeed larger (average 57.6% organisms > 4cm) than in the

305 other five floodplains (from 0.3% in the Rhône to 38.0% in the Tagliamento). Moreover, an average

306 80.9% of the organisms in the Paraná had long life spans compared to less than 40% in temperate,

307 Mediterranean, and arid river floodplains. In accordance with other studies from tropical and

308 subtropical environments (Leigh and Sheldon 2009, Zilli et al. 2008), feeding habits of taxa from the

309 Paraná had high percentages of filter feeders (26.3%), deposit feeders (50.1%), and predators (19.2%). In

310 contrast, shredders were less abundant in the Paraná (1.4%), compared to the other river floodplains

311 (from 10.7% in the Rhône to 39.36% in the Murray, Table 4). This result agrees with previous findings

312 that shredders prefer cool waters and temperate zone riparian vegetation that produce more palatable

313 and nutritious leaves compared with tropical environments (Yule et al. 2009). Furthermore, previous

314 study on aquatic macroinvertebrates in the Paraná River found a general paucity of shredders (Ezcurra

315 de Drago et al. 2007).

316 In contrast, trait categories that were negatively related to the first FCA axis discriminated temperate

317 rivers (upper left quadrat) and included: organisms with spiracle or plastron respiration, short life span,

318 terrestrial or vegetation clutches oviposition by intermediate-to-small body sizes (< 1 cm),

319 predominantly scrapers and piercers feeding on plant material (Fig. 3). Most of these characteristics

320 were in accordance with the trait profile reported by Statzner and Bêche (2010) for lotic

321 macroinvertebrates in Europe and North America. Small differences – for instance in respiration method

322 and reproduction — can be explained by the inclusion of both lotic and lentic habitats in our study

323 compared to Statzner and Bêche (2010)’s work which focused only on lotic environments. Arid and

324 Mediterranean floodplains, located mostly at negative coordinates of both first and second FCA axes

325 (lower left corner), were related to ovoviviparous reproduction, shredding and filter feeding, and

326 intermediate-to-large body sizes (2 to 4 cm and > 8 cm) (Fig. 3). In these environments, shredders,

12

327 scrapers, and deposit feeders were the most abundant groups, accounting for 65-67% of the

328 assemblages (Table 4). Finally, the variance explained by the hydrological connectivity according to the

329 between-class analysis was again very low though significant at 4.0% (P < 0.001).

330

331 In accordance with our predictions, overlap among the six rivers in these multivariate analyses was

332 higher for the trait-based compared to the taxonomic analysis, suggesting greater similarity among

333 climatically disparate river floodplains in biological traits compared to taxonomic composition. These

334 results agree with various studies reporting high trait stability in macroinvertebrate assemblages across

335 Europe (Statzner and Bêche 2010, Statzner et al. 2001, Statzner et al. 2004). They are also congruent

336 with Bonada et al. (2007) who found trait variability between temperate and Mediterranean rivers to be

337 significantly lower than taxonomic variability. These results confirm that trait composition can be a

338 useful tool to compare community patterns across broad large scales where taxonomic composition

339 strongly differs. Finally, differences among hydrological connectivity categories accounted for only 4% of

340 the variability in taxon and trait composition. This figure is much lower than expected since many

341 authors – including those that provided data included in this analysis — attribute an important role to

342 hydrological changes for shaping benthic macroinvertebrate assemblages. More direct indicators of the

343 frequency, magnitude, seasonality or physicochemical effects of floods would have probably explained a

344 more relevant percentage of taxa and trait variability (see for example Gallardo et al. 2009b, Paillex et

345 al. 2007). As demonstrated by Statzner and Bêche (2010), the more directly a stressor acts on a

346 particular trait, the less equivocal the interpretation of that trait response should be.

347

348 Benthic macroinvertebrate patterns across lateral connectivity gradients

349 We used linear mixed effects models (LME) to test our initial hypotheses with all available data while

350 accounting for among river floodplain variability. Although there was considerable diversity among

351 systems, as illustrated by ordination techniques described above, LME allowed us to test whether any of

352 our hypotheses about macroinvertebrate distributions across lateral connectivity gradients were valid at

353 the intercontinental scale. At least 7 out of our 15 initial predictions were confirmed by LME, which

13

354 highlights the importance of hydrological connectivity in influencing macroinvertebrate patterns in

355 floodplains under any climate (Table 5).

356

357 Taxonomic structure within rivers

358 Total macroinvertebrate richness peaked at intermediate levels of hydrological connectivity (Fig. 4a),

359 thus potentially supporting the Intermediate Disturbance Hypothesis (Connell 1978), assuming the

360 connectivity gradient equates to disturbance. As highlighted by Leigh and Sheldon (2009), although the

361 response of richness to connectivity was initially described for temperate rivers (e.g. Arscott et al. 2005,

362 Tockner and Ward 1999, Ward et al. 1999), it has been found to be appropriate for describing

363 biodiversity patterns in Mediterranean (e.g. Gallardo et al. 2009b), arid (e.g. Leigh and Sheldon 2009,

364 Marshall et al. 2006, Sheldon et al. 2002), subtropical and tropical systems (e.g. Leigh and Sheldon 2009,

365 Thomaz et al. 2007, Zilli et al. 2008). As predicted, EPT and COH richness had inverse patterns, peaking

366 respectively at most connected and fragmented floodplain sites (Fig. 4c-d). These patterns were in

367 accordance with an increase in Ephemeroptera, Plecoptera and Trichoptera taxa and a decrease in

368 Coleoptera and Odonata reported along a gradient of lateral hydrological connectivity in the Rhône

369 (Paillex et al. 2009). Apart from macroinvertebrates, hydrological connectivity has been found to modify

370 the richness of many other aquatic organisms such as aquatic plants (e.g. Amoros and Bornette 1999),

371 fishes (e.g. Sheaves et al. 2007), zooplankton (e.g. Frisch et al. 2005, Gallardo et al. 2009b) and

372 phytoplankton (e.g. Gallardo et al. 2009b, van den Brink et al. 1994). Moreover, Tockner et al. (1999)

373 and later Arscott et al. (2005) demonstrated that different taxonomic groups show overlapping richness

374 optima across a gradient of lateral connectivity in the Danube and Tagliamento rivers respectively, with

375 a maximum overall richness in the Danube occurring at intermediate levels of connectivity. Models

376 using taxonomic diversity were not significantly different among hydrological categories, despite

377 exhibiting a trend towards higher diversity at intermediately connected sites (Fig.4b, Table 4).

378

379 Trait structure within rivers

380 In agreement with previous studies reporting positive relationships between taxonomic and trait

381 richness and diversity at multiple scales (Bêche and Statzner 2009, Gallardo et al. 2011, Heino et al.

14

382 2007), taxon and trait richness and diversity were well correlated (Pearson r=0.64 and r= 0.63 at P<

383 0.001 respectively, see regression lines in Figure 5). Trait richness and diversity tended to peak at

384 intermediate levels of connectivity (Fig. 4 e-f) although only the first one was significantly related to

385 connectivity. LME results suggested the frequency of particular trait categories across the connectivity

386 gradient were uneven (Table 5). We expected a higher and more frequent level of flood disturbance to

387 occur in the main river channel and predicted that disturbance regime favour small sized organisms with

388 short life spans and more than one reproductive cycle per year (plurivoltinism). However, small sized

389 organisms were ubiquitous along the connectivity gradient (average 7-11% in all three categories, Fig.

390 6a), peaking at highly connected sites only in the case of the Tagliamento River (Appendix 3a).

391 Furthermore, plurivoltinism and short life span increased in fragmented waterbodies in most rivers, in

392 contrast to our initial hypotheses (Table 5, Fig. 6b-c). Unexpected patterns in macroinvertebrates life

393 history traits might be related to the use of data from the summer season only, when prolonged

394 fragmentation may impose more intensive abiotic constraints on macroinvertebrates than flooding,

395 hence favouring small plurivoltine and short-lived organisms in fragmented habitats (Sheldon et al.

396 2010). This is congruent with results from arid rivers (Appendix 3 a-c), where macroinvertebrate life

397 history patterns were opposite to those of more permanent temperate rivers, as expected considering

398 the intermittent nature of floodplain fragmented water bodies (Sheldon et al. 2010). To a lower extent,

399 the effect of fragmentation can be also noted in Mediterranean and subtropical environments

400 (Appendix 3a-c). Had we included more direct surrogates of hydrological disturbance, or data from a

401 complementary season (e.g. spring), we might have obtained different results, which highlight the need

402 to consider seasonality when evaluating the spatio-temporal heterogeneity of river floodplain.

403 The proportion of individuals with traits that confer resistance to desiccation (diapause, statoblasts, and

404 cocoons) were not significantly different across floodplains when considered together (Table 5, Fig. 6d),

405 but they were more numerous at low connectivity sites in subtropical and arid river floodplains

406 (Appendix 3d). Diapause was the most common resistance strategy (average 19.3%) followed by

407 statoblasts (average 5.3%) and cocoons (average 4.7%), but the majority of taxa had no resistance at all

408 to desiccation (average 68.0%). Other studies have elucidated spatial patterns in resistance to

409 desiccation, for example, across the Paraná River, with greater densities of oligochaetes (able to resist

15

410 water loss through diapause or cysts) in marginal temporary fluvial wetlands (Montalto and Marchese

411 2005). As another example, Reckendorfer (2006) found higher densities of molluscs that utilize aerial

412 respiration and have minimal water loss in low connectivity sites that experience anoxic periods and

413 drought in the Danube river floodplain. Nonetheless, and in accordance with our study, in Statzner and

414 Bêche (2010)’s review of European and North American benthic macroinvertebrates, the majority

415 (>60%) of taxa had no desiccation resistance strategy.

416 Among the feeding habits, the frequency of shredder and scraper were significantly higher in connected

417 sites (Fig. 6e-f), which may be related to the abundance of coarse organic matter and attached algae

418 providing food (Statzner and Bêche 2010). In contrast, deposit feeders were more abundant in

419 fragmented water bodies were organic matter tends to accumulate (Fig. 6g). Interestingly, the

420 proportion of shredders was low in the subtropical system, where leaf decomposition is primarily driven

421 by microbial action rather than by macroinvertebrates (Wantzen and Wagner 2006). Zilli et al. (2008)

422 pointed out that many of the shredder taxa present in northern rivers are absent in the Paraná, where

423 they are mainly represented by chironomids and decapods. Filter feeders were more frequent at

424 intermediate levels of connectivity (Fig. 6h), probably related to the availability of suspended and

425 colloidal particles and lower shear stress currents compared to main channels. Finally, predators are

426 usually large generalist species and were thus expected to show a high number of individuals in

427 fragmented sites, which was confirmed in temperate and Mediterranean river floodplains (Appendix 3i).

428 This pattern was nevertheless reversed in subtropical and arid rivers, where predators were more

429 abundant and dominated highly connected sites, congruent with patterns in body size (large) and life

430 history (long life spans, one reproductive cycle per year) described before. This may be again related to

431 the temporality of fragmented habitats in arid rivers that eventually desiccate, not allowing for the

432 survival of more long-lived or univoltine organisms; whereas permanence of highly connected channels

433 allows the colonization of larger specialist predators belonging, in the case of arid rivers, to Decapoda

434 (e.g. Macrobrachium sp.), Hirudinea (e.g. Erpobdella sp.), Diptera (e.g. Chaoborus sp.), Coleoptera (e.g.

435 Enochrus sp.) and Odonata (e.g. Austrogomphus sp.) families.

436

16

437 In summary, macroinvertebrate patterns in temperate and Mediterranean rivers generally agreed with

438 our a priori hypotheses, while trait differences along connectivity gradients in subtropical and arid rivers

439 were either not significant or in opposition to our expectations. The extent to which global

440 macroecological factors (i.e. climate, dispersal history) and local biotic and abiotic factors (i.e. drought

441 frequency, habitat structure, water chemistry) contributed to this difference cannot be determined with

442 the available data, which highlights the need of more international coordinated studies using

443 comparable protocols for site selection, sampling and analysis. Apart from the effect of water

444 permanency already mentioned, human influence may greatly disrupt the response of

445 macroinvertebrates to hydrological connectivity by reducing the frequency of large erosive floods that

446 normally reset aquatic habitats, or by changing the seasonality of high and low flows. As a way of

447 example, in the Ebro River, discharge has decreased by 30% since 1980s, embankments and dykes have

448 counteracted floodplain interaction, and irrigation of lowland areas has resulted in abnormal high

449 groundwater tables in summer (Cabezas et al. 2009b). Such hydrological disruptions are common to

450 other large regulated rivers, dramatically affecting the spatio-temporal heterogeneity of the floodplain,

451 hence the functional response of macroinvertebrates.

452

453 Data considerations

454 The response of benthic macroinvertebrates to lateral connectivity gradients can vary along a given river

455 corridor. Those responses are primarily linked to hydro-geomorphological differences among floodplain

456 reaches that influence habitat heterogeneity (physical, chemical, and thermal; e.g. Arscott et al. 2000)

457 and inundation dynamics. In the Tagliamento River, Arscott et al. (2005) observed significant changes in

458 benthic macroinvertebrate assemblages across a lateral gradient from headwaters to lowland stretches

459 of the river, mainly associated with changes in the floodplain spatio-temporal heterogeneity and the

460 scarcity of lentic habitats upstream. Although data in our study correspond to mid and lowland reaches

461 in all six river floodplains, differences among rivers in meandering and braiding patterns and floodplain

462 heterogeneity may be influencing our observed responses. In addition, had we included more

463 hydrological categories in this analysis, more complex and perhaps more robust patterns could have

464 been detected. Furthermore, hydrological connectivity modulates invertebrate patterns through

17

465 multiple changes in their habitat (e.g. substrate and vegetation structure, water chemistry), food

466 availability (e.g. transport and accumulation of organic matter and dissolved nutrients) and biotic

467 interaction (e.g. transport of species through active and passive drift). We would probably have found

468 clearer patterns in invertebrate functionality had we used these types of direct factors instead of a

469 ranked estimate of hydrological connectivity (see examples using direct hydrological indicators in

470 Gallardo et al. 2009a, Gallardo et al. 2009b, Sheldon and Thoms 2006).

471 Sampling methodology, taxonomic resolution, and trait scoring may have also influenced our results.

472 However, we did not observe distinct clustering of data related to sampling methodology in ordination

473 plots. Chironomidae and Oligochatea are certainly important components of the riverine aquatic

474 communities that are rarely identified to genus level. In the Paraná River more than 14 different

475 Oligochaeta and 20 Chironomidae genera were identified (Zilli and Marchese 2011), while in arid rivers

476 authors reported at least 22 Chironomidae genera (Sheldon and Thoms 2006), which would have

477 notably increased richness scores in this study. In addition to increasing total richness, identification of

478 these two groups may provide information on adaptation of communities to variable levels of

479 hydrological connectivity. In spite of this potential, Juget and Lafont (1994) reported very little

480 agreement with their quantification of the Rhône River’s habitat templet after identifying more than 50

481 species of Oligochaeta. More accurate trait scores for non-temperate macroinvertebrates are still

482 necessary. In this study we assigned family average trait scores to those uncoded taxa in Tachet’s (2010)

483 classification (Appendix 2). In addition, real traits measured in the field can greatly vary from one region

484 to another, and may even deviate from potential traits represented in Tachet et al. (2010). Thus

485 comparison between rivers from different climates would undoubtedly benefit from accurate and

486 comparable ecological information available for all species.

487

488 Conclusions and perspectives

489 In our study, we compared the taxonomic and trait composition of benthic macroinvertebrate

490 assemblages from a broad representation of river floodplains under different climates. We adopted a

491 macroecological approach to investigate the influence of climate and hydrological connectivity on

492 macroinvertebrate community patterns, paying less attention to the inherent complexity in river

18

493 communities –described in detail by authors listed in Table 2. Instead, we focused on broad spatial scale

494 patterns in an effort to determine comparability across climatic zones. Statistical analyses revealed

495 common taxonomic and trait changes in macroinvertebrate assemblages, which may help disentangle

496 complex responses in assemblages to different degrees of hydrological connectivity.

497 We assumed that trait patterns would be more consistent than taxonomic patterns across large

498 biogeographic scales because the trait filters across floodplain environments may be similar, whereas

499 taxonomic response would likely depend on historical and evolutionary factors (Bonada et al. 2006).

500 Indeed, there was considerably more overlap among river floodplain sites in a multivariate ordination

501 based on trait composition compared to a taxonomy based one. We interpret this as trait stability across

502 hydrological connectivity and suggest that it provides a meaningful ecological context for benthic

503 macroinvertebrate comparison among climate zones, when taxonomic composition strongly differs.

504 Across lateral connectivity gradients, most of our taxonomic specific hypotheses were validated when

505 incorporating data from all six river floodplains (e.g., EPT, COH, and total richness differences across

506 floodplains). We also tested predictions of changes in trait categories across hydrologic conditions that

507 were proposed by Townsend and Hildrew (1994) as part of the River Habitat Templet (RHT). We

508 assumed that in river floodplains, connected sites experience greater changes in habitat conditions

509 related to flood disturbance (i.e. greater change in connectivity over time) and thus experience larger

510 temporal variability. By contrast, fragmented sites have more constant water level and temporal

511 variability is relatively muted. As a result, maximum body size and longevity should decrease across this

512 connectivity gradient, and reproduction frequency, proportion of filter feeders, and scrapers and

513 shredders should increase. Likewise, a range of variable body sizes, life spans, and reproduction

514 preferences may be unconstrained in fragmented and isolated habitats. However, extreme periods of

515 fragmentation may reverse these patterns in arid rivers. Despite the solid logic behind the RHT, we

516 found difficulties testing these predictions. Other authors have also reported significant relationships

517 between invertebrate traits and habitat utilization, though failed to support the RHT predictions (see

518 Juget and Lafont 1994, Resh et al. 1994, Richoux 1994, Usseglio-Polatera and Tachet 1994 and other

519 papers in the same volume). To explain such disagreement, Resh et al. (1994) suggested that traits occur

520 as alternative suites of adaptive characteristics in various groups of organisms and show complex trade-

19

521 offs, and that it is easier to test ecological predictions using individual groups of taxa rather than a more

522 complete macroinvertebrate assemblage. Trait similarities observed in our study suggest that large scale

523 filters still operate, resulting in apparently different trait combinations in temperate and Mediterranean

524 river floodplains when compared to arid and tropical environments. This emphasizes the effect of

525 climate, and especially the interplay of floods and droughts, as primary drivers of macroinvertebrate

526 traits, complexity that may have been underestimated in the RHT. Consequently, studies based on traits

527 should be cautious when using a priori hypotheses extracted from the RHT, and consider the particular

528 climatic and hydrogeomorphological determinants of the floodplain’s spatio-temporal heterogeneity to

529 develop ecologically meaningful predictions.

530

531 Acknowledgements

532 Research in the Ebro River was funded by the Spanish Ministry of Science and Innovation (MICINN

533 CGL2008 – 05153-C02-01/BOS) and supported by the Departments of Environment and Research-

534 Aragon Government (Research Group E-61 on Ecological Restoration). The main author B.G. received

535 funding from the Department of Science Technology and University- Aragon Government (B061-2005

536 pre-doctoral grant) and CAI-Foundation (Programa Europa XXI, travel fellowship). Data from the Rhône

537 River were obtained during the ecological and hydraulic restoration program of the upper Rhône River

538 funded by the Compagnie Nationale du Rhône (CNR), Agence de l’Eau Rhône-Méditerranée et Corse,

539 Diren Délégation de Bassin (RMC), Région Rhône-Alpes and Syndicat du Haut-Rhône. Research in the

540 Paraná River was funded by Consejo Nacional de Investigaciones Científicas y Técnicas (CONICET) and

541 Universidad Nacional del Litoral (UNL).

542

543 Literature cited

544 Amoros, C. and Bornette, G. 1999. Antagonistic and cumulative effects of connectivity: a predictive 545 model based on aquatic vegetation in riverine wetlands. — Arch Hydrobiol 311-327. 546 Amoros, C. and Bornette, G. 2002. Connectivity and biocomplexity in waterbodies of riverine 547 floodplains. — Freshwater Biol 47: 761-776. 548 Amoros, C. and Roux, A. L. 1988. Interaction between large water bodies within the floodplain of large 549 rivers: function and development of connectivity. — Münstersche Geographische Arbeiten 29: 550 125-130.

20

551 Arrington, D. A. and Winemiller, K. O. 2006. Habitat affinity, the seasonal flood pulse, and community 552 assembly in the littoral zone of a Neotropical floodplain river. — J N Am Benthol Soc 25: 126- 553 141. 554 Arscott, D. B. et al. 2000. Aquatic habitat diversity along the corridor of an Alpine floodplain river (Fiume 555 Tagliamento, Italy). — Arch Hydrobiol 149: 679-704. 556 Arscott, D. B. et al. 2005. Lateral organization of aquatic invertebrates along the corridor of a braided 557 floodplain river. — J N Am Benthol Soc 24: 934-954. 558 Bêche, L. A. and Statzner, B. 2009. Richness gradients of stream invertebrates across the USA: 559 taxonomy- and trait-based approaches. — Biodivers Conserv 18: 3909-3930. 560 Bolker, B. M. et al. 2009. Generalized linear mixed models: a practical guide for ecology and evolution. 561 — Trends Ecol Evol 24: 127-135. 562 Bonada, N. et al. 2007. Taxonomic and biological trait differences of stream macroinvertebrate 563 communities between mediterranean and temperate regions: implications for future climatic 564 scenarios. — Global Change Biology 13: 1658-1671. 565 Bonada, N. et al. 2006. Benthic macroinvertebrate assemblages and macrohabitat connectivity in 566 Mediterranean-climate streams of northern California. — J N Am Benthol Soc 25: 32-43. 567 Botta-Dukat, Z. 2005. Rao's quadratic entropy as a measure of functional diversity based on multiple 568 traits. — Journal of Vegetation Science 16: 533-540. 569 Bunn, S. E. et al. 2003. Sources of organic carbon supporting the food web of an arid zone floodplain 570 river. — Freshwater Biol 48: 619-635. 571 Cabezas, A. et al. 2009a. Hydrologic and landscape changes in the Middle Ebro River (NE Spain): 572 implications for restoration and management. — Hydrol Earth Syst Sc 13: 1-12. 573 Cabezas, A. et al. 2009b. The effect of anthropogenic disturbance on the hydrochemical characteristics 574 of riparian wetlands at the Middle Ebro River (NE Spain). — Hydrobiologia 617: 101-116. 575 Castella, E. et al. 1984. Macroinvertebrates as describers of morphological and hydrological types of 576 aquatic ecosystems abandoned by the Rhône River. — Hydrobiologia 119: 219-225. 577 Champely, S. and Chessel, D. 2002. Measuring biological diversity using Euclidean metrics. — 578 Environmental and Ecological Statistics 9: 167-177. 579 Charvet, S. et al. 2000. Traits of benthic macroinvertebrates in semi-natural French streams: an initial 580 application to biomonitoring in Europe. — Freshwater Biol 43: 277-296. 581 Chessel, D. et al. 2004. The ade4 package. I: One-table methods. — R News 4: 5-10. 582 Chevenet, F. et al. 1994. A fuzzy coding approach for the analysis of long-term ecological data. — 583 Freshwater Biol 31: 295-309. 584 Connell, J. H. 1978. Diversity in tropical rain forests and coral reefs - high diversity of trees and corals is 585 maintained only in a non-equilibrium state. — Science 199: 1302-1310. 586 Depetris, P. J. 2007. The Parana River under extreme flooding: A hydrological and hydro-geochemical 587 insight. — Interciencia 32: 656-662. 588 Depetris, P. J. et al. 1996. ENSO-controlled flooding in the Parana River (1904 - 1991). — 589 Naturwissenschaften 83: 127-129. 590 Dolédec, S. and Chessel, D. 1989. Seasonal successions and spatial variables in freshwater environments. 591 2. Distinction between seasonal and environmental-effects within floral and faunistical data. — 592 Acta Oecologica-Oecologia Generalis 10: 207-232. 593 Dolédec, S. et al. 1999. Species traits for future biomonitoring across ecoregions: patterns along a 594 human-impacted river. — Freshwater Biol 42: 737-758. 595 Drago, E. C. 1989. Morphological and hydrological characteristics of the floodplain ponds of the Middle 596 Parana River, Argentina. — Revue d'Hydrobiologie Tropicale 22: 183-190. 597 Drago, E. C. et al. 2003. Aquatic habitats, fish amd invertebrate assemblages of the Middle Paraná River. 598 — Amazoniana 17: 291-341. 599 Dray, S. and Dufour, A. B. 2007. The ade4 package: implementing the duality diagram for ecologists. — 600 Journal of Statistical Software 22: 1-20. 601 Dray, S. et al. 2007. The ade4 package - II: Two-table and K-table methods. — R News 7: 47-52. 602 Ezcurra de Drago, I. et al. 2007. Benthic Invertebrates. — In: Iriondo, M. H. et al. (eds), The Middle 603 Paraná River: Limnology of a Subtropical Wetland. Springer-Verlag. 604 Frisch, D. et al. 2005. Diversity of floodplain copepods (Crustacea) modified by flooding: species 605 richness, diapause strategies and population genetics. — Arch Hydrobiol 162: 1-17.

21

606 Gallardo, B. et al. 2008. Macroinvertebrate patterns along environmental gradients and hydrological 607 connectivity within a regulated river-floodplain. — Aquat Sci 70: 248-258. 608 Gallardo, B. et al. 2009a. Testing the response of the macroinvertebrate functional structure and 609 biodiversity to flooding and confinement. — Journal of Limnology 68: 315-326. 610 Gallardo, B. et al. 2011. How to choose a biodiversity indicator - Redundancy and complementarity of 611 biodiversity metrics in a freshwater ecosystem. — Ecological Indicators 11: 1177-1184. 612 Gallardo, B. et al. 2009b. Modelling the response of floodplain aquatic communities across the lateral 613 hydrological gradient. — Mar Freshwater Res 60: 924-935. 614 Gasith, A. and Resh, V. H. 1999. Streams in Mediterranean climate regions: Abiotic influences and biotic 615 responses to predictable seasonal events. — Annu Rev Ecol Syst 30: 51-81. 616 Gayraud, S. et al. 2003. Invertebrate traits for the biomonitoring of large European rivers: an initial 617 assessment of alternative metrics. — Freshwater Biol 48: 2045-2064. 618 Heino, J. et al. 2007. Ecological filters and variability in stream macroinvertebrate communities: do 619 taxonomic and functional structure follow the same path? — Ecography 30: 217-230. 620 Hill, M. O. 1974. Correspondence Analysis - Neglected multivariate method. — Journal of the Royal 621 Statistical Society Series C-Applied Statistics 23: 340-354. 622 Juget, J. and Lafont, M. 1994. Theoretical habitat templets, species traits, and species richness - aquatic 623 oligochaetes in the Upper Rhône River and its floodplain. — Freshwater Biol 31: 327-340. 624 Junk, W. J. et al. 1989. The flood pulse concept in river-floodplain systems. — Canadian Special 625 Publications of Fisheries and Aquatic Sciences 106: 110-127. 626 Junk, W. J. and Wantzen, K. M. 2004. The flood pulse concept: new aspects, approaches, and 627 applications. An update. . — In: Welcomme, R. and Petr, T. (eds), Proceedings of the 2nd Large 628 River Symposium (LARS). Food and Agriculture Organization & Mekong River Commission, pp. 629 117-149. 630 Kennard, M. J. et al. 2010. Classification of natural flow regimes in Australia to support environmental 631 flow management. — Freshwater Biol 55: 171-193. 632 Keppel, G. 1991. Design and analysis: a researcher handbook. — Prentice Hall. 633 Knighton, A. D. and Nanson, G. C. 1994. Waterholes and their significance in the anastomosing channel 634 system of Cooper Creek, Australia. — Geomorphology 9: 311-324. 635 Leigh, C. and Sheldon, F. 2009. Hydrological connectivity drives patterns of macroinvertebrate 636 biodiversity in floodplain rivers of the Australian wet /dry tropics. — Freshwater Biol 54: 549- 637 571. 638 Marchese, M. and Ezcurra de Drago, I. 1992. Benthos of the lotic environments in the Middle Parana 639 River system - transverse zonation. — Hydrobiologia 237: 1-13. 640 Marchese, M. et al. 2002. Benthic macroinvertebrates and physical habitat relationships in the Parana 641 River flood-plain system. — In: IAHS, S. p. (ed), The Ecohydrology of South American Rivers and 642 Wetlands. 643 Marshall, J. C. et al. 2006. The macroinvertebrate fauna of an Australian dryland river: spatial and 644 temporal patterns and environmental relationships. — Mar Freshwater Res 57: 61-74. 645 Merritt, R. W. et al. 2002. Wetland macroinvertebrates of Prentiss Bay, Lake Huron, Michigan: Diversity 646 and functional group composition. — Great Lakes Entomologist 35: 149-160. 647 Minshall, G. W. et al. 1983. Interbiome comparison of stream ecosystem dynamics. — Ecol Monogr 53: 648 1-25. 649 Moilanen, A. and Nieminen, M. 2002. Simple connectivity measures in spatial ecology. — Ecology 83: 650 1131-1145. 651 Montalto, L. and Marchese, M. 2005. Cyst formation in Tubificidae (Naidinae) and Opistocystidae 652 (Annelida, Oligochaeta) as an adaptive strategy for drought tolerance in fluvial wetlands of the 653 Parana River, Argentina. — Wetlands 25: 488-494. 654 Oksanen, J. et al. 2007. vegan: Community ecology package. — In: http://cran.r-project.org/ (ed). 655 Olivier, J. M. et al. 2009. The Rhône River Basin. — In: Tockner, K. et al. (eds), Rivers of Europe. Academic 656 Press, pp. 247-296 657 Paillex, A. et al. 2007. Aquatic macroinvertebrate response along a gradient of lateral connectivity in 658 river floodplain channels. — J N Am Benthol Soc 26: 779-796. 659 Paillex, A. et al. 2009. Large river floodplain restoration: predicting species richness and trait responses 660 to the restoration of hydrological connectivity. — J Appl Ecol 46: 250-258.

22

661 Pearson, R. G. and Boyero, L. 2009. Gradients in regional diversity of freshwater taxa. — J N Am Benthol 662 Soc 28: 504-514. 663 Peel, M. C. et al. 2007. Updated world map of the Koppen-Geiger climate classification. — Hydrol Earth 664 Syst Sc 11: 1633-1644. 665 Petts, G. and Amoros, C. 1993. Fluvial Hydrosystems. — Chapmann & Hall. 666 Pinheiro, J. et al. 2010. nlme: Linear and nonlinear mixed effects models. 667 Poff, N. L. and Ward, B. 1990. Physical Habitat Templet of lotic systems: recovery in the context of 668 historical pattern of spatio temporal heterogeneity. — Environmental Management 14: 629- 669 645. 670 Poff, N. L. and Ward, J. V. 1989. Implications of Streamflow Variability and Predictability for Lotic 671 Community Structure - a Regional-Analysis of Streamflow Patterns. — Can J Fish Aquat Sci 46: 672 1805-1818. 673 Puckridge, J. T. et al. 1998. Flow variability and the ecology of large rivers. — Mar Freshwater Res 49: 55- 674 72. 675 R Development Core Team 2011. R: A language and environment for statistical computing. — R 676 Foundation for Statistical Computing 677 Reckendorfer, W. et al. 2006. Floodplain restoration by reinforcing hydrological connectivity: expected 678 effects on aquatic mollusc communities. — The Journal of applied ecology 43: 474-484. 679 Resh, V. H. et al. 1994. Theoretical habitat templets, species traits, and species richness - A synthesis of 680 long-term ecological research on the Upper Rhône River in the context of concurrently 681 developed ecological theory. — Freshwater Biol 31: 539-554. 682 Richoux, P. 1994. Theoretical habitat templets, species traits, and species richness - aquatic Coleoptera 683 in the Upper Rhône River and its floodplain. — Freshwater Biol 31: 377-395. 684 Sheaves, M. et al. 2007. Fish fauna of dry tropical and subtropical estuarine floodplain wetlands —Mar 685 Freshwater Res 58: 931-943. 686 Sheldon, F. et al. 2002. Conservation value of variable connectivity: aquatic invertebrate assemblages of 687 channel and floodplain habitats of a central Australian arid-zone river, Cooper Creek. — Biol 688 Conserv 103: 13-31. 689 Sheldon, F. et al. 2010. Ecological roles and threats to aquatic refugia in arid landscapes: dryland river 690 waterholes. — Mar Freshwater Res 61: 885-895. 691 Sheldon, F. and Thoms, M. C. 2006. Relationships between flow variability and macroinvertebrate 692 assemblage composition: Data from four Australian dryland rivers. — River Res Appl 22: 219- 693 238. 694 Skern, M. et al. 2010. Aquatic Heteroptera as indicators for terrestrialisation of floodplain habitats. — 695 Limnologica 40: 241-250. 696 Statzner, B. and Bêche, L. A. 2010. Can biological invertebrate traits resolve effects of multiple stressors 697 on running water ecosystems? — Freshwater Biol 55: 80-119. 698 Statzner, B. et al. 2001. Perspectives for biomonitoring at large spatial scales: a unified measure for the 699 functional composition on invertebrate communities in European running waters. — Basic and 700 Applied Ecology 2: 73-85. 701 Statzner, B. et al. 2004. Biological trait composition of European stream invertebrate communities: 702 assessing the effects of various trait filter types. — Ecography 27: 470-488. 703 Tachet, H. et al. 2010. Invertebrés d'eau douce: Systématique, Biologie, Écologie. — CNRS éditions. 704 ter Braak, C. J. F. 1986. Canonical Correspondence-Analysis - A new eigenvector yechnique for 705 multivariate direct gradient analysis. — Ecology 67: 1167-1179. 706 Thomaz, S. M. et al. 2007. Floods increase similarity among aquatic habitats in river-floodplain systems. 707 — Hydrobiologia 579: 1-13. 708 Tockner, K. et al. 2000. An extension of the flood pulse concept. — Hydrol Process 14: 2861-2883. 709 Tockner, K. et al. 1999. The Danube restoration project: Species diversity patterns across connectivity 710 gradients in the floodplain system. — Regul River 15: 245-258. 711 Tockner, K. and Ward, J. V. 1999. Biodiversity along riparian corridors. — Arch Hydrobiol 293-310. 712 Tockner, K. et al. 2003. The Tagliamento River: a model ecosystem of European importance. — Aquat Sci 713 65: 239-253. 714 Townsend, C. R. and Hildrew, A. G. 1994. Species traits in relation to a habitat templet for river systems. 715 — Freshwater Biol 31: 265-275.

23

716 Usseglio-Polatera, P. et al. 2000. Biomonitoring through biological traits of benthic macroinvertebrates: 717 how to use species trait databases? — Hydrobiologia 422: 153-162. 718 Usseglio-Polatera, P. and Tachet, H. 1994. Theoretical habitat templets, species traits, and species 719 richness - Plecoptera and Ephemeroptera in the Upper Rhône River and its floodplain. — 720 Freshwater Biol 31: 357-375. 721 van den Brink, F. W. B. et al. 1994. Impact of hydrology on phytoplankton and zooplankton community 722 composition in floodplain lakes along the Lower Rhine and Meuse. — J Plankton Res 16: 351- 723 373. 724 Walker, K. F. and Thoms, M. C. 1993. Environmental-effects of flow regulation on the lower river 725 Murray, Australia. — Regul River 8: 103-119. 726 Wantzen, K. M. et al. 2008. Ecological effects of water-level fluctuations in lakes: an urgent issue. — 727 Hydrobiologia 613: 1-4. 728 Wantzen, K. M. and Wagner, R. 2006. Detritus processing by invertebrate shredders: a neotropical- 729 temperate comparison. — J N Am Benthol Soc 25: 216-232. 730 Ward, J. V. and Stanford, J. A. 1995. Ecological connectivity in alluvial river ecosystems and its disruption 731 by flow regulation. — Regul River 11: 105-119. 732 Ward, J. V. et al. 1999. Biodiversity of floodplain river ecosystems: Ecotones and connectivity. — Regul 733 River 15: 125-139. 734 Yule, C. M. et al. 2009. Shredders in Malaysia: abundance and richness are higher in cool upland tropical 735 streams. — J N Am Benthol Soc 28: 404-415. 736 Zilli, F. L. et al. 2008. Benthic invertebrate assemblages and functional feeding groups in the Parana River 737 floodplain (Argentina). — Limnologica 38: 159-171. 738 Zilli, L. F. and Marchese, R. M. 2011. Patterns in macroinvertebrate assemblages at different spatial 739 scales. Implications of hydrological connectivity in a large floodplain river. — Hydrobiologia 740 663: 245-257.

24

741 Table 1. Expected response of macroinvertebrate metrics (taxonomic and trait) among floodplains in different climate regions (A), and within floodplains along a gradient of

742 lateral hydrological connectivity (B). EPT: Ephemeroptera, Plecoptera and Trichoptera. COH: Coleoptera, Odonata and Heteroptera.

Scale Metrics Type Expectation Reference Taxa Taxon Differences among climate regions Poff and Ward 1990

composition

(A) Trait Trait Differences less pronounced than taxa composition Charvet et al. 2000 composition Bonada et al. 2007 EPT Taxon Peak in well oxygenated habitats with a high hydrological connectivity because of the high dissolved oxygen level Usseglio-Polatera and richness requirement and lower thermal tolerances of associated species Tachet 1994

COH richness Taxon Peak at fragmented sites because of life-history habits (large size, aerial respiration, predation) and habitat association Bonada et al. 2006 (shoreline vegetation) Skern et al. 2010 Taxonomic Taxon Peak at intermediate levels of hydrological connectivity where generalist and specialist species coexist following the Connell 1978 richness and intermediate disturbance hypothesis diversity Trait richness Trait Peak in intermediate and fragmented sites where hydrologic stability may foster biotic interaction Statzner et al. 2004

and diversity

(B) Body size* Trait Small size (< 0.5 cm) allows organisms resisting near-bottom flow stress in connected sites Townsend and Hildrew 1994 Statzner and Bêche 2010 Number of Trait Plurivoltinism (> 1 cycle per year) allows resilience to frequent flood disturbance in highly connected sites Townsend and Hildrew 1994 cycles/year* Reference to arid Life cycle Trait In close relation to plurivoltinism, short life span duration allows resilience to frequent flood disturbance in highly Townsend and Hildrew 1994 duration* connected sites Resistance Trait Resistance forms such as cocoons, cells against desiccation or diapause allow resistance to desiccation in fragmented Townsend and Hildrew 1994 forms sites Feeding habits Trait Filterer feeders, scrapers and shredders are adapted to efficiently exploit limited food resources in highly connected Merritt et al. 2002 sites, including coarse particulate matter, attached algae, and suspended particles Gallardo et al. 2009a Stable intermediate to fragmented sites may favour large specialist predators, and deposit feeders feeding on organic Statzner and Bêche 2010 matter accumulations

743 *Patterns in arid rivers may be the opposite because of abiotic constraints related to prolonged periods of extended hydrological fragmentation, i.e. higher proportion of organisms with small 744 body size, plurivoltinism and short life span under low hydrological connectivity.

25

745 Table 2. River floodplains investigated in this study. N= number of sampling points within each river. More information on the connectivity of sites and methods used to

746 sample macroinvertebrates can be found in references cited in the Benthic macroinvertebrate sampling column.

Climate-River N High connectivity Intermediate connectivity Low connectivity Sampling season Benthic macroinvertebrate sampling Subtropical 53 Site in a secondary Two shallow wetlands with Two shallow lakes November 2005 to Nine samples per wetland and three in the Paraná channel (Miní river) relatively high hydrological with low February 2006 secondary channel. Macroinvertebrates sampled connection with the main connection to the with an Ekman grab (0.0225 m2) and sieved through channel main channel a 200 µm mesh (Zilli and Marchese 2011) Temperate 42 Three sites along the Three backwaters connected Six fragmented August 1998 Data obtained from lowland reaches IV, V and VI Tagliamento main river channel downstream only wetlands near or described in Arscott et al. (2005) and three secondary far from the main Macroinvertebrates collected in triplicate from channels channel each site with a Hess sampler (0.04 m2; 100 µm mesh size) Temperate 28 Two sites in the main Six backwaters connected Nine lentic isolated July 2003 Average values from quadruple quadrat samples Rhône river channel and two downstream only water bodies 0.25 m2 per site, macroinvertebrates sampled with secondary channels a hand net, 500 µm mesh (Paillex et al. 2007) Mediterranean 43 Three sites along the Four backwaters connected La Alfranca, Cartuja August 2006 Triplicate samples per site. Macroinvertebrates Ebro main river channel downstream only and Juslibol sampled with a hand net (500 µm mesh size) and two secondary isolated oxbow covering a 0.25 m2 surface (Gallardo et al. 2009a) channels lakes Arid 20 Six permanent and Five temporary water bodies Nine ephemeral December 1991 Samples collected from most abundant Cooper semi-permanent water bodies Cooper, and microhabitats (e.g. large woody debris, aquatic water bodies along November 1993 vegetation and unvegetated areas) by sweeping a the main channel Diamantina rivers 500 µm mesh pond net over 5 m2 for 20 seconds Arid 20 Seven sites connected Seven semi-permanent Six ephemeral December 1990 (Sheldon et al. 2002, Sheldon and Thoms 2006) Murray upstream and (flooded > 75% of the year) or (flooded <10% downstream, temporal (flooded 50-75% of year) isolated permanently flooded the year) sites connected up or oxbow lakes downstream only

26

747 Table 3. Biological traits and categories for benthic macroinvertebrates considered in the present study 748 (extracted from Tachet et al. 2010). 749

Trait Categories Trait Categories Maximum size < 0.25 cm Resistance Eggs, statoblasts, 0.25 < size ≤ 0.5 cm form Cocoons 0.5 < size ≤1 cm Cells against desiccation 1 < size ≤ 2 cm Diapause or dormancy 2 < size ≤ 4 cm None 4 < size ≤ 8 cm Flier > 8 cm Locomotion Surface swimmer Respiration Tegument Swimmer Gill Burrower Plastron Crawler Spiracle (aerial) Interstitial Hydrostatic vesicle Temporarily attached Life cycle duration < 1year (short life span) Permanently attached > 1 year (long life span) Food Fine sediments and Potential number of < 1 (univoltine) microorganisms rep. cycles/yr 1 (semivoltine) Detritus < 1 mm > 1 (plurivoltine) Plant detritus ≥ 1 mm Aquatic stage Egg Living microphytes Larva Living macrophytes Nymph Dead animal > 1 mm Adult Living microinvertebrates Reproduction Ovoviviparity Living macroinvertebrates Isolated eggs, cemented Vertebrates Clutches, cemented or Feeding Deposit feeders fixed habits Filter feeders Asexual reproduction Shredders Isolated eggs, free Scrapers Clutches, free Piercers Eggs or clutches in Predators vegetation Parasites Dispersal Aquatic passive Aquatic active Aerial passive Aerial active 750

751

752

27

753 Table 4. Mean (± s.d.) of taxonomic and trait attributes for the six floodplains investigated. Values are

754 the mean of all the sites in each floodplain. Differences among rivers are all significant (Kruskal-Wallis

755 test, df=5, P < 0.01).

Metrics Type Paraná Tagliamento Rhône Ebro Murray Cooper

Taxonomic Taxon 7.30 13.50 29.89 5.79 16.15 16.80

Richness (±2.37) (±6.53) (±9.30) (±2.52) (±4.11) (±5.69)

Taxonomic Taxon 1.39 0.96 1.63 0.96 1.46 1.56

Diversity (± 0.51) (± 0.34) (± 0.54) (± 0.46) (± 0.35) (± 0.33)

EPT Richness Taxon 8.50 12.48 14.16 21.27 8.30 12.05

(±13.03) (±21.18) (±19.15) (±37.50) (±9.89) (±20.86)

COH Richness Taxon 0.00 0.59 11.97 12.76 17.47 23.83

(±0.00) (±1.04) (±16.56) (±29.63) (±30.51) (±28.23)

Trait Richness Trait 40.60 53.02 56.25 48.28 57.90 59.20

(±5.57) (±3.40) (±1.76) (±4.98) (±2.13) (±1.79)

Trait Diversity Trait 58.84 41.30 48.05 44.10 47.41 57.51

(±19.16) (±17.66) (±13.81) (±19.20) (±15.17) (±19.36)

Size <0.5 cm Trait 3.48 1.48 9.74 14.78 17.22 21.78

(%) (±3.75) (±1.74) (±9.48) (±26.18) (±21.41) (±19.10)

Short life span Trait 19.14 60.26 70.71 66.94 61.86 65.44

(%) (±17.37) (±36.33) (±12.47) (±28.74) (±23.54) (±24.97)

Plurivoltinism Trait 61.14 79.47 53.42 59.99 40.02 50.31

(%) (±18.33) (±16.03) (±5.86) (±25.92) (±21.42) (±24.33)

Resistance Trait 32.47 28.49 41.07 30.68 24.98 34.55

forms (%) (±13.24) (±11.25) (±16.80) (±23.04) (±21.27) (±16.90)

Deposit feeders Trait 50.14 54.24 16.31 30.38 11.83 20.89

(%) (±23.70) (±27.37) (±7.86) (±24.12) (±5.80) (±17.78)

Shredders Trait 1.35 10.72 23.40 29.86 39.36 28.79

(%) (±2.62) (±6.46) (±14.04) (±16.61) (±14.37) (±8.21)

28

Scrapers Trait 0.74 12.25 17.17 14.83 15.99 14.72

(%) (±1.53) (±12.23) (±11.35) (±10.12) (±9.06) (±12.21)

Filter feeders Trait 26.33 9.99 12.89 10.30 6.36 5.33

(%) (±22.23) (±8.18) (±10.29) (±10.87) (±2.98) (±3.90)

Predators Trait 19.23 8.75 9.20 7.67 18.11 16.84

(%) (±17.00) (±10.11) (±7.29) (±7.23) (±14.68) (±10.97)

756

757

29

758 Table 5. Results from Linear Mixed Effects models (LME) performed on 15 taxonomic and functional

759 indicators using floodplain identity as random factor and hydrological connectivity

760 (High/Intermediate/Low) as fixed factor. Corr: correlation between observed and fitted values. n.s. not

761 significant. Trend observed: hydrological category where the metric shows a highest score. Accordance:

762 support (YES) or rejection (NO) of expectations described in Table 1.

763

Metrics Type Corr F2,198 P-value Trend Trend Accordance

expected observed

Taxon. Richness Taxon 0.84 7.15 0.001 Intermediate Intermediate YES

Taxon. Diversity Taxon 0.53 1.23 n.s. Intermediate none

EPT Richness Taxon 0.66 30.36 0.001 High High YES

COH Richness Taxon 0.80 2.34 0.099 Low Low YES

Trait Richness Trait 0.59 2.74 0.066 Inter. to Low Intermediate YES

Trait Diversity Trait 0.37 0.01 n.s. Inter. to Low none

Size < 0.5 cm Trait 0.42 0.91 n.s. High none

Short life span Trait 0.64 0.09 0.091 High Low NO

Plurivoltinism Trait 0.52 3.29 0.039 High Low NO

Resistance form Trait 0.26 0.14 n.s. Low none

Shredders Trait 0.79 10.00 0.001 High High YES

Scrapers Trait 0.61 9.52 0.001 High High YES

Filter feeders Trait 0.57 11.07 0.001 High Intermediate NO

Deposit feeders Trait 0.62 3.30 0.039 Low Low YES

Predators Trait 0.42 2.03 n.s. Low none

764

30

765 FIGURES

766 Figure 1. Location of the six river floodplains included in this study: subtropical (Paraná), temperate

767 (Tagliamento and Rhône), Mediterranean (Ebro) and arid (Murray and Cooper). Ombrothermic diagrams

768 (temperature: red line and rainfall: blue dashed line) are shown to illustrate the main climatic

769 differences among regions.

770 Figure 2. Results of a Correspondence Analysis (CA) performed on taxon composition in the six studied

771 river floodplains. Ellipses envelop 1.5 times variance of observations for each floodplain (dots represent

772 sampling points). Taxa showing the highest correlation scores with each axis are given (BIV: Bivalvia,

773 COL: Coleoptera, DIP: Diptera, EPH: Ephemeroptera, PLE: Plecoptera, TRI: Trichoptera).

774 Figure 3. Results of a Fuzzy Correspondence Analysis (FCA) performed on trait composition in six river

775 floodplains. Ellipses envelop 1.5 times variance of observations for each floodplain. Trait categories

776 showing the highest correlation scores with each axis are given. See Table 3 for further information on

777 trait categories.

778 Figure 4. Boxplots of macroinvertebrate taxonomic metrics across lateral hydrological connectivity

779 gradients (High, Intermediate and Low). Statistics given on the right upper corner correspond to a Linear

780 Mixed Effects (LME) model using ‘river identity’ as random factor and ‘hydrological connectivity’ as fixed

781 factor. Degrees of freedom of all models=198, n.s.= non-significant model at P > 0.1.

782 Figure 5. Relationship between taxon and trait richness (a); and between taxon and trait diversity (b).

783 Data from 6 rivers under three different climates pulled together. The best regression line (logarithmic

784 or linear) and corresponding equation are shown.

785 Figure 6. Boxplot of nine macroinvertebrate traits across hydrological connectivity gradients (High,

786 Intermediate and Low). Statistics given on the right upper corner correspond to a Linear Mixed Effects

787 (LME) model using ‘river identity’ as random factor and ‘hydrological connectivity’ as fixed factor.

788 Degrees of freedom of all models=198, n.s.= non-significant model at P> 0.1.

31

789 Figure 1

790

32

791 Figure 2

792

33

793 Figure 3

794

795

34

796 Figure 4

797

35

798 Figure 5

799

36

800 Figure 6

801

802

37

803 Appendix 1. Taxa list of macroinvertebrates and their presence (+) in the six large floodplain rivers

804 investigated. P= Paraná, T= Tagliamento, R= Rhône, E= Ebro, M= Murray, C= Cooper.

River Order Class Genus/Family P T R E M C Bivalvia Unionoida Anodonta + Unionoida Unio + + Unionoida Velesunio + Veneroida Corbicula + + Veneroida Corbiculina + Veneroida Dreissena + Veneroida Eupera + Veneroida Musculium + Veneroida Pisidium + Veneroida Sphaeriidae + + + Gastropoda Archaeogastropoda Theodoxus + + Basommatophora Acroloxus + Basommatophora Ancylus + Basommatophora Anisus + Basommatophora Ferrissia + + + + Basommatophora Galba + + Basommatophora Glyptophysa + Basommatophora Gyraulus + + Basommatophora Heleobia + Basommatophora Hippeutis + Basommatophora Isidorella + + Basommatophora Lymnaea + Basommatophora Physa + + + + Basommatophora Physella + Basommatophora Planorbis + Basommatophora Radix + Basommatophora Stagnicola + + Heterostropha Valvata + Mesogastropoda Bithynella + Mesogastropoda Notopala + Mesogastropoda Potamopyrgus + Mesogastropoda Thiara + Mesogastropoda Viviparus + Hirudinea Arhynchobdellida Erpobdella + + Hirudinea Haementeria + Rhynchobdellida Batracobdella + Rhynchobdellida Glossiphonia + + Rhynchobdellida Helobdella + Rhynchobdellida Hemiclepsis + Rhynchobdellida Piscicola + Rhynchobdellida Theromyzon + Insecta Coleoptera Allodessus + + Coleoptera Anacaena + Coleoptera Berosus + + + Coleoptera Bidessus Coleoptera Brychius + Coleoptera Colymbetinae + Coleoptera Copelatus + Coleoptera Coxelmis + Coleoptera Cybister + Coleoptera Dryops + + Coleoptera Dytiscus + Coleoptera Elmis + +

38

Coleoptera Enochrus + + Coleoptera Eretes + Coleoptera Esolus + + Coleoptera Graptodytes + Coleoptera Gyrinus + Coleoptera Haliplus + + Coleoptera Helochares + + Coleoptera Hydaticus + Coleoptera Hyderodes + Coleoptera Hydraena + + + + Coleoptera Hydrobius + Coleoptera Hydrochara + Coleoptera Hydroglyphus + Coleoptera Hydroporus + Coleoptera Hydrovatus + + Coleoptera Hygrotus + Coleoptera Hyphydrus + + Coleoptera Ilybius + Coleoptera Laccobius + + Coleoptera Laccophilus + + Coleoptera Limnius + + Coleoptera Limnoxenus + + Coleoptera Macrogyrus + Coleoptera Megaporus + + Coleoptera Nebrioporus + Coleoptera Necterosoma + Coleoptera Noterus + Coleoptera Ochthebius + + + + Coleoptera Orechtochilus + + Coleoptera Oulimnius + Coleoptera Paracymus + + Coleoptera Paroster + Coleoptera Peltodytes + Coleoptera Platambus + Coleoptera Pomatinus + Coleoptera Rhantus + + Coleoptera Riolus + Coleoptera Scarodytes + + Coleoptera Sternopriscus + Coleoptera Stictotarsus + Hemiptera Agraptocorixa + + Hemiptera Antiporus + + Hemiptera Corixa + + Hemiptera Cymatia + Hemiptera Enithares + Hemiptera Gerris + + + Hemiptera Hydrometra + Hemiptera Ilyocoris + Hemiptera Mesovelia + + Hemiptera Micronecta + + + + Hemiptera Microvelia + + Hemiptera Naucoris + + + Hemiptera Nepa + Hemiptera Notonecta + Hemiptera Plea + Hemiptera Ranatra + Hemiptera Sigara + Odonata Aeshna + + Odonata Anax + Odonata Austrogomphus + 39

Odonata Calopteryx + Odonata Coenagrion + + + Odonata Cordulia + Odonata Crocothermis + Odonata Erythromma + Odonata Gomphus + Odonata Hemicordulia + + Odonata Ischnura + + Odonata Lestes + Odonata Libellula + + Odonata Nehalennia + Odonata Onychogomphus + Odonata Orthetrum + Odonata Platycnemis + Odonata Pseudagrion + Odonata Pyrrhosoma + Odonata Sympetrum + Odonata Trithemis + Odonata Xanthagrion + + Diptera Atherix + Diptera Bezzia + + Diptera Blepharicera + Diptera Ceratopogoninae + + + + + Diptera Chaoborus + + Diptera Chironomidae + + + + + + Diptera Clinocerinae + + + Diptera Culicinae + + + + Diptera Dolichopodidae + Diptera Eloeophila + Diptera Ephydridae + + + + Diptera Eriopterini + Diptera Hexatomini + + Diptera Limoniini + + + Diptera Molophilus + Diptera Muscidae + + + Diptera Pilaria + Diptera Psychodidae + Diptera Ptychopteridae + Diptera Chrysophylum + Diptera Sciomyzidae + Diptera Simuliidae + + Diptera Stratiomyidae + + + Diptera Tabanidae + + Diptera Tipulidae + + + + + Megaloptera Sialis + + Ephemeroptera Atalophlebia + Ephemeroptera Baetis + + + Ephemeroptera Caenis + + + Ephemeroptera Campsurus + Ephemeroptera Centroptilum + Ephemeroptera Cloeon + + + + + Ephemeroptera Ecdyonurus + + + Ephemeroptera Epeorus + Ephemeroptera Ephemera + + + Ephemeroptera Habroleptoides + + Ephemeroptera Heptagenia + + Ephemeroptera Potamanthus + Ephemeroptera Rhithrogena + + Ephemeroptera Serratella + + Ephemeroptera Siphlonurus + + 40

Ephemeroptera Tasmanocoenis + + Plecoptera Amphinemura + Plecoptera Chloroperla + Plecoptera Dinocras + Plecoptera Isoperla + Plecoptera Leuctra + + Plecoptera Nemoura Plecoptera Perla + + Plecoptera Protonemura + + Trichoptera Anabolia + Trichoptera Athripsodes + Trichoptera Beraeodes + Trichoptera Ceraclea + Trichoptera Chaetopteryx + Trichoptera Cheumatopsyche + Trichoptera Cyrnus + Trichoptera Ecnomus + + + Trichoptera Glossosoma + + + Trichoptera Glyphotaelius + Trichoptera Goera + Trichoptera Halesus + Trichoptera Hydropsyche + + Trichoptera Hydroptila + + Trichoptera Lepidostoma + + Trichoptera Leptocerus + + Trichoptera Limnephilus + + Trichoptera Mystacides + Trichoptera Neureclipsis + Trichoptera Oecetis + + + Trichoptera Orthotrichia + Trichoptera Philopotamus + Trichoptera Phryganea + Trichoptera Polycentropus + Trichoptera Psychomyia + + Trichoptera Rhyacophila + + Trichoptera Sericostoma + Trichoptera Setodes + Trichoptera Silo + Trichoptera Tinodes + Trichoptera Triplectides + + Trichoptera Triplectides + Malacostraca Amphipoda Crangonyx + Amphipoda Echinogammarus + + Amphipoda Gammarus + + Amphipoda Niphargus + + Decapoda Atyaephyra + Decapoda Caridina + Decapoda Cherax + + Decapoda Macrobrachium + + + Decapoda Orconectes + Decapoda Paratya + + Decapoda Procambarus + + Isopoda Asellus + + Isopoda Proasellus + + Oligochaeta + + + + + + 805

806

41

807 Appendix 2. Taxa not found in Tachet et al. (2010) trait classification, and the taxon used as reference

808 for trait analysis. When a reference taxon could not be identified, the average of the family was used.

Uncoded taxa Alternative reference Uncoded taxa Alternative reference Agraptocorixa Corixidae> Corixa sp. Limnephilus Limnephilidae> other Limnephilini Allodesus Hydroporinae (average) Macrobrachium Cambarus sp. Antiporus Hydroporinae (average) Macrogyrus Gyrinidae> Gyrinus sp. Aphylla Gomphiidae (average) Megaporus Dytiscidae> Hydroporinae (average) Atalophlebia Leptophlebiidae (average) Molophilus Tipulidae> Eriopterini (average) Austragrion Coenagrionidae (average) Musculium Corbiculidae> Corbicula sp. Bezzia Ceratopogonidae (average) Necterosoma Dytiscidae> Hydroporinae Campsurus Polymitarcyidae> Ephoron sp. Notopala Viviparidae> Viviparus sp. Caridina Atyidae> Atyaephira sp. Octhebius Hydraenidae> Hydraena sp. Chaetopterix Limnephilidae> other Paratya Atyidae> Atyaephira sp. Chaetopterygini Cherax Parastacidae> Astacus sp. Paroster Dytiscidae> Hydroporinae Coxelmis Elmidae> Elmis sp. Pillaria Tipulidae> Hexatomini (average) Enithares Notonectidae> Notonecta sp. Pseudagrion Coenagrionidae> Coenagrion sp. Eupera Sphaeriidae> Psidium sp. Sternopriscus Dytiscidae> Hydroporinae sp. Glyptophysa Physidae> Physa sp. Tasmanocoenis Caenidae> Caenis sp. Heleobia Hydrobiidae (average) Triplectides Leptoceridae> Mystacides sp. Hemicordulia Corduliidae> Cordulia sp. Trithemis Libellulidae (average) Hyderodes Dytiscidae> Dytiscus sp. Velesunio Unionidae> Unio sp. Isidorella Bullinidae> type Planorbis sp. Xantagrion Coenagrionidae> Coenagrion sp. 809

810

42

811 Appendix 3. Boxplots of macroinvertebrate traits across a gradient of lateral hydrological connectivity

812 (High, Intermediate and Low) in each of the six river floodplains investigated. Chi-sq and P-values

813 correspond to non-parametric analysis of variance (Kruskal Wallis test). n.s.: not significant at P>0.1.

814

815

816

43

817 Appendix 3 (cont.)

818

819

820

821

44